This document describes exploratory and some detailed data analysis

Sample map

Map of sediment sample locations at Ashfield Flats for all sampling years 2019-2022.

Figure 1: Map of sediment sample locations at Ashfield Flats for all sampling years 2019-2022.

Samples were taken each year from 2019 to 2022 (Figure 1). Different sampling designs were implemented each year depending on the perceived gaps in understanding of the site and the learning objectives desired for the students involved. Apart from three soil/sediment depth profiles in 2019, all samples were subaqueous or subaerial surface sediments (0-10cm) from stormwater drains, wetland ponds (with or without water), saltmarsh, or seasonally flooded woodland.

Data summaries

pH, EC, Al-Cu


Fe-Ni


P-Zn

 

The elements of primary interest in this study are the rare-earth (REE or lanthanide) elements La, Ce, Nd, and Gd; Y (often considered together with the REE); major elements considered to be useful proxies for sediment parameters expected to affect geochemical reactions of REE: Al, Ca, Fe, P and S. Aluminium (Al) is a proxy for clay minerals (although pXRD data show that other aluminosilicates are present, such as feldspars and micas, these are resistant and less likely than the phyllosilicate clays to report Al to an aqua regia digest). Calcium (Ca) is a proxy for carbonate minerals (since silicate-bound Ca would also be resistant to dissolution in aqua regia). Iron (Fe) concentrations are a proxy for iron oxide (and possibly iron sulfide) minerals. Phosphorus (P) is included since secondary REE phosphates such as rhabdophane may be a REE sink. Sulfur (S) most likely represents sulfate and sulfide minerals, would be expected to accumulate in wet, reducing environments, and is strongly linked to acid sulfate soils.

Sediment pH and EC have numerous but not excessive missing observations and are included due to their substantial effects on sediment geochemical processes.

The trace elements As, Cu, Pb, and Zn are included as common inorganic contaminants. In addition, Pb may be immobilised in environments receiving acid sulfate soil drainage due to the insolubility of PbSO4. Thorium (Th) may be depleted in oxidised acid sulfate soils.

Note that several elements have too many missing observations to be useful: Cd, Ga, Rb, Sc, and Ti.


Drain sediment


Lake sediment


Saltmarsh


Distribution tests

Zn

0.428

7.06e-28

0.967

1.19e-05

0.967

1.23e-05

-0.00973

0

0.0243

0.372


Most variables except pH are not normally distributed (Table 7). No variables have normal distributions when log10-transformed. A few variables have normal distributions when power-transformed: pH, Ba, & Ca. The non-normal distributions of variables, even when transformed, suggest non-parametric tests are required e.g. for means comparisons).

check Na bimodality

This is worth doing to see if there is more than one population of samples based on salinity (assuming the main source of Na to aqua regia digests is halite).

The resulting map (Figure 2) shows a smaller population of low-Na locations, corresponding with where low salinity would be expected (i.e. away from tidal influence, and/or where evaporative concentration of Na salts is unlikely).

Density histograms for potentially multi-modal variables.

Figure 2: Density histograms for potentially multi-modal variables.

Additional explanatory maps

Surface water wetland naming

Map of wetland pond locations at Ashfield Flats comparing naming systems.

Figure 3: Map of wetland pond locations at Ashfield Flats comparing naming systems.

Sampling zones

Map of sampling zones at Ashfield Flats 2019-2022

Figure 4: Map of sampling zones at Ashfield Flats 2019-2022

Means by Kruskal-Wallis

Comparing means (strictly mean rank sums) between sampling zones (the factor ZoneSimp) as this seems most appropriate. As shown in Table 8, all overall effects are significant at p<0.001. Some post-hoc pairwise comparisons show significant differences (P$$0.05), differing for each variable.

REE boxplot by Zone

Sum of REE concentrations by sampling zone at Ashfield Flats for all sampling years. Different colours show environment sub-types, and means are different (p<0.05, Kruskal-Wallis) if italic text below x-axis labels has no common letters.

Figure 5: Sum of REE concentrations by sampling zone at Ashfield Flats for all sampling years. Different colours show environment sub-types, and means are different (p<0.05, Kruskal-Wallis) if italic text below x-axis labels has no common letters.

 

The greatest concentrations of REE are in the north and northeast wetland pond sediments (SW05 (N) and SW06 (NE) in McGrath 2021, see Figure 5). The pattern of REE means across zones is similar to that for Al in Figure 6.

Aluminium boxplot by Zone

Al concentrations by sampling zone at Ashfield Flats for all sampling years. Different colours show environment sub-types, and means are different (p<0.05, Kruskal-Wallis) if italic text below x-axis labels has no common letters.

Figure 6: Al concentrations by sampling zone at Ashfield Flats for all sampling years. Different colours show environment sub-types, and means are different (p<0.05, Kruskal-Wallis) if italic text below x-axis labels has no common letters.

Bubble maps

∑REE

Map of REE concentrations by location at Ashfield Flats for all sampling years.

Figure 7: Map of REE concentrations by location at Ashfield Flats for all sampling years.

Al

Map of Al concentrations by location at Ashfield Flats for all sampling years.

Figure 8: Map of Al concentrations by location at Ashfield Flats for all sampling years.

Boxplots by Type

∑REE and Al, Ca, Fe, S

Boxplots of REEs and selected major and trace elements by sample type at Ashfield Flats for all sampling years (DS = Drain sediment; Fl = Flooded; LS = Lake Sediment; SM = Saltmarsh; So = Soil).

Figure 9: Boxplots of REEs and selected major and trace elements by sample type at Ashfield Flats for all sampling years (DS = Drain sediment; Fl = Flooded; LS = Lake Sediment; SM = Saltmarsh; So = Soil).

Boxplots by Zone

Ce, La, Nd, Gd

Boxplots of REEs by sampling Zone at Ashfield Flats for all sampling years (DS = Drain sediment; Fl = Flooded; LS = Lake Sediment; SM = Saltmarsh; So = Soil).

Figure 10: Boxplots of REEs by sampling Zone at Ashfield Flats for all sampling years (DS = Drain sediment; Fl = Flooded; LS = Lake Sediment; SM = Saltmarsh; So = Soil).

Al, Ca, Fe, S

Boxplots of selected major elements by sampling Zone at Ashfield Flats for all sampling years.

Figure 11: Boxplots of selected major elements by sampling Zone at Ashfield Flats for all sampling years.

Co, Cr, Cu, Ni, Pb, Zn

Boxplots of selected major elements by sampling Zone at Ashfield Flats for all sampling years (CMD = Chapman Drain, KMD = Kitchener Drain, LP = Limestone path, N = North wetland pond, NE = North-east wetland pond, NW = North-west wetland pond, S = South wetlands / side drain, SE = South-east wetland pond, SM = Saltmarsh (east of CMD), SW = South-west wetland pond, WD = Woolcock Drain).

Figure 12: Boxplots of selected major elements by sampling Zone at Ashfield Flats for all sampling years (CMD = Chapman Drain, KMD = Kitchener Drain, LP = Limestone path, N = North wetland pond, NE = North-east wetland pond, NW = North-west wetland pond, S = South wetlands / side drain, SE = South-east wetland pond, SM = Saltmarsh (east of CMD), SW = South-west wetland pond, WD = Woolcock Drain).

Scatter plot matrix majors & REEs raw

Scatter plot matrix for log10-transformed elements at Ashfield Flats for all sampling years.

Figure 13: Scatter plot matrix for log10-transformed elements at Ashfield Flats for all sampling years.

Scatter plots

Ce, La, Nd, Gd vs. Al

REE-Al relationships at Ashfield Flats for all sampling years.

Figure 14: REE-Al relationships at Ashfield Flats for all sampling years.

Ce, La, Nd, Gd vs. Fe

REE-Fe relationships at Ashfield Flats for all sampling years.

Figure 15: REE-Fe relationships at Ashfield Flats for all sampling years.

Ce, La, Nd, Gd vs. P

REE-P relationships at Ashfield Flats for all sampling years.

Figure 16: REE-P relationships at Ashfield Flats for all sampling years.

## Untransformed:
##      Al   Ca    Fe   P     S   La   Ce   Nd   Gd   Cu   Pb   Th   Zn
## 1  9000 5180 28800 630 32400  8.4 11.4  4.7 1.24 19.0 31.1  2.1 4180
## 2 37100 3280 44700 544 17900 48.6 93.0 36.6 6.51 75.8 43.2 15.7 1220
## 3 28700 4430 48600 419 13900 41.5 81.6 32.5 5.73 54.7 32.0 15.3  418
## 4 33900 2840 53100 533  9880 41.9 81.0 33.1 5.90 62.9 41.8 18.6  270
## 5 31200 3050 71500 744  8540 41.5 82.9 34.9 6.90 50.5 34.7 12.4  173
## 6 34800 4220 38400 465  8910 35.4 72.5 29.0 5.39 58.3 36.9 16.0  176
## 
## CLR-transformed:
##     Al   Ca   Fe     P    S    La     Ce    Nd    Gd     Cu     Pb    Th    Zn
## 1 4.70 4.15 5.86 2.040 5.98 -2.28 -1.970 -2.86 -4.19 -1.460 -0.969 -3.66 3.930
## 2 5.28 2.86 5.47 1.060 4.55 -1.36 -0.708 -1.64 -3.37 -0.913 -1.480 -2.49 1.860
## 3 5.20 3.34 5.73 0.978 4.48 -1.33 -0.658 -1.58 -3.31 -1.060 -1.590 -2.33 0.976
## 4 5.23 2.75 5.68 1.080 4.00 -1.47 -0.806 -1.70 -3.43 -1.060 -1.470 -2.28 0.396
## 5 5.31 2.99 6.14 1.580 4.02 -1.31 -0.619 -1.48 -3.10 -1.110 -1.490 -2.52 0.116
## 6 5.47 3.36 5.57 1.160 4.11 -1.42 -0.702 -1.62 -3.30 -0.920 -1.380 -2.21 0.186

Scatter plot matrix majors-REEs CLR-transformed

Scatter plot matrix of CLR-transformed elements at Ashfield Flats for all sampling years.

Figure 17: Scatter plot matrix of CLR-transformed elements at Ashfield Flats for all sampling years.

Scatter plot matrix of CLR-transformed elements at Ashfield Flats for all sampling years.

Figure 18: Scatter plot matrix of CLR-transformed elements at Ashfield Flats for all sampling years.

Scatter Plots for CLR-transformed variables

Ce, La, Nd, Gd vs. Al

REE-Al relationships (concentrations CLR-transformed) at Ashfield Flats for all sampling years.

Figure 19: REE-Al relationships (concentrations CLR-transformed) at Ashfield Flats for all sampling years.

Principal components analysis

PCA Summary

## Importance of components:
##                           PC1    PC2    PC3     PC4     PC5    PC6     PC7
## Standard deviation     2.8922 1.4590 1.4110 1.19798 1.02381 0.9101 0.82625
## Proportion of Variance 0.4403 0.1120 0.1048 0.07554 0.05517 0.0436 0.03593
## Cumulative Proportion  0.4403 0.5523 0.6571 0.73263 0.78780 0.8314 0.86733
##                            PC8     PC9    PC10    PC11    PC12    PC13    PC14
## Standard deviation     0.72077 0.71313 0.67446 0.51658 0.44520 0.41887 0.36647
## Proportion of Variance 0.02734 0.02677 0.02394 0.01405 0.01043 0.00923 0.00707
## Cumulative Proportion  0.89467 0.92144 0.94538 0.95942 0.96986 0.97909 0.98616
##                          PC15    PC16    PC17    PC18    PC19
## Standard deviation     0.3515 0.24809 0.18308 0.17264 0.12075
## Proportion of Variance 0.0065 0.00324 0.00176 0.00157 0.00077
## Cumulative Proportion  0.9927 0.99590 0.99766 0.99923 1.00000

Contribution of variables to principal components

PCA Biplots by Type

bp12 <- fviz_pca_biplot(af_pca, geom="point",col.ind = data0$Type,
                            title = "Dimensions 1, 2", axes = c(1,2)) +
  theme_bw()
bp23 <- fviz_pca_biplot(af_pca, geom="point",col.ind = data0$Type,
                            title = "Dimensions 2, 3", axes = c(2,3)) +
  theme_bw()
ggarrange(bp12,bp23,ncol=2)

Principal components biplot for 2019-2022 Ashfield data grouped by sample Type.

Figure 20: Principal components biplot for 2019-2022 Ashfield data grouped by sample Type.

PCA Biplots by Zone

bp12 <- fviz_pca_biplot(af_pca, col.ind = data0$Zone, title="PC1 & PC2 by Zone",
                        geom="point", lwd = 2, axes = c(1,2), 
                        palette=rainbow(nlevels(data0$Zone),v=0.7,end=0.75)) +
  theme_bw()
bp23 <- fviz_pca_biplot(af_pca, col.ind = data0$Zone, title="PC2 & PC3 by Zone",
                        geom="point", lwd=2, axes = c(2,3), 
                        palette=rainbow(nlevels(data0$Zone),v=0.7,end=0.75)) +
  theme_bw()
ggarrange(bp12, bp23, nrow = 1)

Principal components biplot for 2019-2022 Ashfield data grouped by sampling Zone.

Figure 21: Principal components biplot for 2019-2022 Ashfield data grouped by sampling Zone.

PCA Biplots by ZoneSimp

bp12 <- fviz_pca_biplot(af_pca, col.ind = data0$ZoneSimp, 
                        title="PC1 & PC2 by ZoneSimp",
                        geom="point", lwd=2, axes = c(1,2), 
                        palette=rainbow(nlevels(data0$ZoneSimp),v=0.7,end=0.75)) +
  theme_bw()
bp23 <- fviz_pca_biplot(af_pca, col.ind = data0$ZoneSimp, 
                        title="PC2 & PC3 by ZoneSimp",
                        geom="point", lwd=2, axes = c(2,3), 
                        palette=rainbow(nlevels(data0$ZoneSimp),v=0.7,end=0.75)) +
  theme_bw()
ggarrange(bp12, bp23, nrow = 1)

Principal components biplot for 2019-2022 Ashfield data grouped by simplified sampling Zone.

Figure 22: Principal components biplot for 2019-2022 Ashfield data grouped by simplified sampling Zone.

References

McGrath, G. S. (2021). Ashfield Flats Hydrological Study: Summary Report. Kensington, Western Australia, Department of Biodiversity, Conservation, and Attractions (Government of Western Australia). http://www.dbca.wa.gov.au/sites/default/files/2021-12/Ashfield%20Flats%20summary%20report.pdf

end of file